Skip to main content
Performance-to-Video models re-render a real filmed take into a new world: your uploaded performance drives the motion — movement, timing, blocking, expressions — while your styled first frame drives the look. In Automat Studio, these models power Performance mode in Shot Studio.
Takes must be 2–15 seconds; output duration is matched to your take. The take’s original audio — your actors’ real voices and on-set sound — can be carried into the generated video with the “Keep my take’s audio” toggle.
Video credits scale with clip duration (2–15 seconds, matched to your take). Keeping the take’s audio costs nothing extra.

​
Kling 3.0 Motion

  • Credits: 39–284
  • Rating: ⭐⭐⭐⭐ (4/5)
  • Provider: fal.ai
  • Best for: High-quality performance transfer at the best price (Default model)

​
Kling 3.0 Motion Pro

  • Credits: 51–378
  • Rating: ⭐⭐⭐⭐⭐ (5/5)
  • Provider: fal.ai
  • Best for: Maximum fidelity for complex movement and expressive performances

​
Supported Models

The Seedance 2.0 models reject reference footage containing likenesses of real people (an upstream content policy — including AI-restyled frames of real people). Use them only for stylized or non-human takes; for filmed performances of real actors, use the Kling Motion models above. With Seedance, the audio toggle generates AI audio rather than keeping your take’s sound.

​
Seedance 2.0

  • Credits: 92–681
  • Rating: ⭐⭐⭐⭐⭐ (5/5)
  • Provider: fal.ai
  • Best for: Maximum-quality reference generation for stylized/non-human takes

​
Seedance 2.0 Fast

  • Credits: 74–545
  • Rating: ⭐⭐⭐⭐⭐ (5/5)
  • Provider: fal.ai
  • Best for: Faster Seedance generation for stylized/non-human takes

​
Seedance 2.0 Mini

  • Credits: 45–341
  • Rating: ⭐⭐⭐⭐ (4/5)
  • Provider: fal.ai
  • Best for: Affordable Seedance generation for stylized/non-human takes